On the improvement and acceleration of eigenvalue decomposition in spectral methods using GPUs
ForfatterJohansen, Thomas A. Haugland
The key objectives in this thesis are; the study of GPU-accelerated eigenvalue decomposition in an effort to uncover both benefits and pitfalls, and then to investigate and facilitate a future GPU implementation of the symmetric QR algorithm with permutations. With the current trend of having ever larger datasets both in terms of features and observations, we propose that GPU computation can help ameliorate the temporal penalties incurred by eigendecomposing large matrices. We successfully show the benefits of performing eigendecomposition on GPUs, and also highlight some problems with current GPU implementations. While implementing the QR algorithm on GPU, we discovered that the GPU-based QR decomposition does not explicitly form the orthogonal matrix needed as part of the QR algorithm. Therefore, we propose a novel GPU algorithm for “implicitly” computing the orthogonal matrix Q from the Householder vectors given by the QR decomposition. To illustrate the benefits of our methods, we show that the kernel entropy component analysis algorithm on GPU is two orders of magnitude faster than an equivalent CPU implementation.
ForlagUiT Norges arktiske universitet
UiT The Arctic University of Norway
Følgende lisensfil er knyttet til denne innførselen:
Viser innførsler relatert til tittel, forfatter og emneord.
Schulz, Jörn (Doctoral thesis; Doktorgradsavhandling, 2013-12-18)The use of statistical shape analysis in medical settings has increased during the last decades. This thesis presents contributions to three major topics of statistical shape analysis with application to medical problems. ...
Long-range memory in Earth surface temperatures: spatial scale dependence and land-sea differences Fredriksen, Hege-Beate (Master thesis; Mastergradsoppgave, 2013-05-15)The use of long-range memory models as a description of the noise in Earth surface temperatures has increased the recent years, and as a measure of the persistence for such time series we have the Hurst exponent. It is ...
Maritime Target Detection in Non-homogeneous Sea Clutter Environments based on Single- and Multi-polarization Synthetic Aperture Radar Data Tao, Ding (Doctoral thesis; Doktorgradsavhandling, 2015-12-03)This thesis discusses the subject of maritime target detection based on single- and multi-polarization synthetic aperture radar (SAR) data. The primary objective is to develop an automatic and effective target detection ...